[asterisk-bugs] [JIRA] (DAHLIN-382) Broken build for Linux ≥ 4.7 with netdev support enabled

[Alexei A Smekalkine (JIRA)]

    [ https://issues.asterisk.org/jira/browse/DAHLIN-382?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=251947#comment-251947 ] 

Alexei A Smekalkine edited comment on DAHLIN-382 at 9/6/20 1:01 PM:
--------------------------------------------------------------------

Most of our work is in a closed OE-based repository and the changes are added as a set of patches. I posted most of it right now for you: https://github.com/ikle/dahdi-linux/commits/work — This changeset is being used in production with Linux 4.4 on top of commit 6667f1c8d8fadc29ef95dce4f8b025f38d3d3084 (pre-3.0). (Also builds for Linux 4.19, but not as thoroughly tested at the moment.) In this case, I did a rebase on next branch of upstream (without conflicts).

* commit fa2addaad403c63b426bed4938b5055d88871bfd (uncomment netdev flag definition) seems to have been added by mistake a long time ago when trying to port version 2.11 to new kernels: I don't see any use of this definition anywhere in the code today. In any case, this change is safe and will not break anything.
* commit a0f79fe3a08f428687c157584c78fd14fbfd8378: add support for hard HDLC netdev — It might make sense to split to two patches and call the open and close methods not only for hard HDLC.
* commit ff1d8ba22c717a3a4901ade5041bf424b957e903: wct4xxp: add support for OpenVox D115/D130 — Needs to be divided to two patches: adding single span devices support and adding OpenVox single span devices support. In the second case, it may be necessary to add a mechanism for loading alternative firmware for OpenVox: We do not have Wildcard devices and we have not tested the ability of OpenVox devices to work with firmware from Wildcard. (OpenVox devices seems to be clone of Wildcard devices.)

Most of the fixes have not yet been submitted here at the moment.

https://github.com/ikle/dahdi-tools/commits/work — Same as for dahdi-linux, used in production.

* Support unframed E1 — nedd support from device driver, as an example: https://github.com/ikle/ds (good sample for DAHDI development and tests: this driver can work without a real hardware in "fake" mode).
* tools/sethdlc: support Cisco HDLC ethernet — support for https://github.com/ikle/ds/blob/master/linux/hdlc_cisco_eth.c, In an amicable way, we need to send it upstream first. Used in production, triggers DAHLIN-381.
* dahdi_cfg: fix HDLC channels aggregation — required for proper PPPoHDLC support (used dchan with pppd to support authentication) and for hard HDLC.

P.S. Let's not litter this task with unrelated comments anymore: use email directly.


was (Author: ikle):
Most of our work is in a closed OE-based repository and the changes are added as a set of patches. I posted most of it right now for you: https://github.com/ikle/dahdi-linux/commits/work — This changeset is being used in production with Linux 4.4 on top of commit 6667f1c8d8fadc29ef95dce4f8b025f38d3d3084 (pre-3.0). (Also builds for Linux 4.19, but not as thoroughly tested at the moment.) In this case, I did a rebase on next branch of upstream (without conflicts).

* commit fa2addaad403c63b426bed4938b5055d88871bfd (uncomment netdev flag definition) seems to have been added by mistake a long time ago when trying to port version 2.11 to new kernels: I don't see any use of this definition anywhere in the code today. In any case, this change is safe and will not break anything.
* commit a0f79fe3a08f428687c157584c78fd14fbfd8378: add support for hard HDLC netdev — It might make sense to split to two patches and call the open and close methods not only for hard HDLC.
* commit ff1d8ba22c717a3a4901ade5041bf424b957e903: wct4xxp: add support for OpenVox D115/D130 — Needs to be divided to two patches: adding single span devices support and adding OpenVox single span devices support. In the second case, it may be necessary to add a mechanism for loading alternative firmware for OpenVox: We do not have Wildcard devices and we have not tested the ability of OpenVox devices to work with firmware from Wildcard. (OpenVox devices seems to be clone of Wildcard devices.)

Most of the fixes have not yet been submitted here at the moment.

https://github.com/ikle/dahdi-tools/commits/work — Same as for dahdi-linux, used in production.

* Support unframed E1 — nedd support from device driver, as an example: https://github.com/ikle/ds.
* tools/sethdlc: support Cisco HDLC ethernet — support for https://github.com/ikle/ds/blob/master/linux/hdlc_cisco_eth.c, In an amicable way, we need to send it upstream first. Used in production, triggers DAHLIN-381.
* dahdi_cfg: fix HDLC channels aggregation — required for proper PPPoHDLC support (used dchan with pppd to support authentication) and for hard HDLC.

> Broken build for Linux ≥ 4.7 with netdev support enabled
> --------------------------------------------------------
>
>                 Key: DAHLIN-382
>                 URL: https://issues.asterisk.org/jira/browse/DAHLIN-382
>             Project: DAHDI-Linux
>          Issue Type: Bug
>      Security Level: None
>          Components: dahdi (the module)
>    Affects Versions: 3.0.0, 3.1.0
>            Reporter: Alexei A Smekalkine
>            Assignee: Keith Morgan
>              Labels: patch
>         Attachments: 0001-dahdi-base-fix-build-with-Linux-4.7-and-netdev-suppo.patch
>
>
> Transaction start time field [removed from net_device structure in Linux 4.7|https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=9b36627acecd5792e81daf1a3bff8eab39ed45fb].
> To fix this use netif_trans_update helper from <linux/netdevice.h>.
> P.S. [The attached patch|^0001-dahdi-base-fix-build-with-Linux-4.7-and-netdev-suppo.patch] has beed tested with Linux 4.19.



--
This message was sent by Atlassian JIRA
(v6.2#6252)